BUVIS - A Novel Approach to Real-Time Automated Visual Inspection
Hulya Yalcin, Mehmet Tezol, Isil Bozma


This report presents a novel system - BUVIS - geared for the real-time automated inspection of industrial parts for post-manufacturing quality control. The aim of the visual inspection is to determine whether all the holes on the part are located correctly and of correct shape within acceptable tolerances while ensuring that the completion time is comparable to that of an human inspector. The novelty of the system is that it is based on an event-driven approach to visual processing and decision-making. In the Phase 1 of this project, an algorithm that selectively processes the incoming visual data - hence being event-driven - and finds all the holes on the given part has been developed. In the second phase of the project, this algorithm has been extended to infer the positions and shapes of all the holes detected. In the final phase of the project, two modes of operation has been added to the system: (i) The learning mode during which a model part is presented to the system which then forms an internal model of the holes and their locations, (ii) the inspection mode during which a ‘to-be-quality-controlled’ part is inspected and compared with the model part to determine its manufacturing quality.

                                                                [ More about this in html... ]

                                                                Download the paper [ps], [pdf]

Resume | Research | Main Page
Carnegie Mellon University, Robotics Institute
5000 Forbes Av., Pittsburgh, PA, 15213
hulya@ri.cmu.edu